A psychiatrist reads a patient’s weekend message at 11 pm. A dermatologist reviews a photo uploaded three days ago. A primary care physician closes out a portal thread that took five exchanges over two weeks to resolve. None of these are visits in the way an AI scribe understands the word, yet all of them need a chart-ready note.
Roughly 40% of US physicians reported using AI clinical documentation tools in 2025, and nearly all of that adoption assumed one thing: a patient and a clinician talking, live, in the same encounter window. Asynchronous medicine AI does not accept that assumption.
The clinical day stopped staying inside the exam room years ago. The documentation tools mostly did not follow it there, and that gap is exactly where this category’s next real product decision sits.
Why Asynchronous Clinical Care Outgrew the Tools Built for It
Patient portals turned physician inboxes into a second clinical shift. One cohort study on protected EHR time found that physicians were already setting aside dedicated appointment slots just to process asynchronous messaging, prescription refills, and prior authorizations, a sign that this work had outgrown the margins of the day.
Async telemedicine vendors built entire businesses around this shift. Fabric Health markets itself as the leading asynchronous telemedicine solution, automating clinical symptom collection and SOAP notes generation across chat, phone, video, and async channels, and claims visits that move 10 times faster than video or in-person care.
That is a real and useful product. It solves intake, routing, and low-acuity triage at scale. It does not address what happens after the note is created: whether the diagnosis and procedure codes that justify reimbursement were ever generated.
Where Real-Time AI Scribes Break Down
The category has spent the past two years optimizing for the live encounter, and it shows. Two distinct failure points emerge once the visit stops being a single recorded conversation.
The Tier 1 Scribes Were Never Built for Fragmented Encounters
Freed, Heidi Health, and Nabla each built strong products around the same core mechanic: listen to one conversation, transcribe it, and structure it into a SOAP note. Heidi positions itself as a “care partner for the full clinical day,” but that promise still assumes a single recorded session. None of the three Tier 1 scribes were designed to ingest a patient’s photo from Tuesday, a messaged symptom update from Thursday, and a follow-up question from the following Monday, then reconstruct a coherent clinical narrative across all three.
DeepScribe comes closer on the billing side, with ICD-10 and HCC coding built into its enterprise oncology product, but at $350 to $500 per month per provider, that depth is priced out of reach for solo and small-group practices that do the bulk of async patient communication today.
Async-Aware Scribes Still Stop at the Note
Empathia AI has moved furthest toward acknowledging the async gap directly, marketing coverage across in-person, telehealth, and asynchronous visits, even without a stable internet connection, while positioning competitors as “real-time only” or restricted to a single EMR add-on. That positioning is closer to the actual problem. But covering async visits and closing the billing loop on them are two different products, and most async-aware scribes stop at the note.
The pattern across the category is consistent. Every vendor markets toward the moment of capture, whether that moment is a 15-minute exam-room conversation or a patient’s photo upload, and treats what happens after capture as a downstream EHR problem rather than part of the product. For a solo family physician working through a backlog of portal messages at the end of the day, that distinction is not academic.
The note still has to carry the correct diagnosis and procedure codes regardless of whether the encounter was live or assembled from three separate touchpoints across a week.
What AI in Asynchronous Medicine Actually Has to Do
Documentation for an asynchronous encounter is reconstruction, not transcription. The physician is not narrating a single conversation for an AI to capture. The system has to pull together fragments collected across days, sometimes across different channels and different staff members, and turn that into a chart-ready note without inventing continuity the patient never provided.
A pilot study comparing ambient-only documentation with documentation enhanced with longitudinal patient history found real value in incorporating prior clinical context rather than capturing a single moment in isolation, which is precisely the design challenge async-native documentation must solve.
The accuracy stakes are not theoretical. The risk is not a generic AI error; it is async-specific: without a single recorded conversation to check against, the physician’s review step becomes the last line of defense, not an optional formality.
Billing Does Not Get Easier When the Visit Is Asynchronous
Asynchronous volume does not just complicate documentation; it complicates the revenue cycle behind it. The breakdown happens in two places.
Fragmented Encounters Make Undercoding Harder to Catch
The internal medicine physician spending hours per day on documentation already loses revenue to undercoding under normal visit volume. Spread that same documentation across a fragmented async thread with no dedicated visit window, and the coding problem compounds. None of the async-first platforms reviewed here, including Fabric Health, is built around ICD-10 and CPT auto-coding as a core capability.
That gap is exactly why Notiro’s three-stage model, patient intake before the encounter, ambient or async-derived documentation during it, and ICD-10/CPT auto-coding after it, was never dependent on a single live recording to begin with. A note assembled from async inputs still needs the same billing-grade coding as a live visit. Most of the category has not been built for that yet.
Two Incomplete Answers Are Not One Complete Workflow
The practice manager weighing async-first vendors against traditional ambient scribes is often forced to choose between two incomplete options. The async platform routes the patient and drafts a note. The scribe accurately captures a live conversation. Neither, on its own, protects the revenue cycle once the visit type no longer resembles a standard 15-minute appointment. As async volume grows across primary care, psychiatry, and dermatology, the documentation tooling that wins will be the one that treats coding as inseparable from the note, not bolted on as an EHR step after the fact.
How Notiro Closes the Gap
Asynchronous care is not a future trend for digital health; it is already where a measurable share of patient-physician communication happens. The platforms built to deliver async visits and the scribes built to document live ones have not yet converged into a single workflow that also protects the billing side of the chart. Undercoding does not become less expensive because the visit was asynchronous; it becomes harder to catch. Notiro’s ICD-10 and CPT auto-coding works from whatever clinical content exists, recorded conversation, or reconstructed async thread, and turns it into a chart-ready, billing-ready note. Start a free trial at Notiro, no IT setup, no enterprise contract.
